维度增量决策表的区分矩阵属性约简方法

Dimension Incremental Attribute Reduction Methods Based on Discernibility Matrix

维度增加是数据动态变化的重要类型之一.为了快速有效地计算这类数据的属性约简,基于区分矩阵提出了两种维度增量决策表的属性约简方法:一种方法是通过新加入属性集的信息更新决策表的区分矩阵,并根据更新后的区分矩阵计算新的约简;另一种方法则是通过更新一种新提出的压缩决策表区分矩阵来计算维度增量后的属性约简.这两种方法都可以获得与非增量约简方法相同的结果,同时还可以显著地降低计算维度增量数据属性约简的耗时,其中基于压缩表区分矩阵计算维度增量数据属性约简的方法更为快速.理论分析和实验结果验证了算法的有效性和高效性.

Dimension increase is one of the important types of dynamic data. In order to calculate attribute reduction of this kind of data efficiently, this paper put forward two types of dimension incremental attribute reduction methods based on discernibility matrix：one way is to compute new reduction by modifying a decision table＇s discernibility matrix;Another way is to compute dimension incre- mental attribute reduction by updating a compacted decision table＇s discernibility matrix. Both methods can obtain the same results with the non-incremental attribute reduction method. They also significantly reduce time-consuming of computing attribute reduction in the case of the dimension of a decision table increasing, and the method based on a compacted decision table is faster. Theoretical anal- ysis and experimental results verify the effectiveness and efficiency of these two proposed algorithms.